Module 4
HIV Testing Strategies
and Algorithms
Purpose
To provide the participants with a basic understanding of how
HIV rapid tests are selected for the country’s algorithm and how
HIV status is determined following the algorithm.
Pre-requisite
Modules
Module 1: Overview of HIV Infection
Learning
Objectives
At the end of this module, you will be able to:
Content Outline
Module 3: Overview of HIV Testing Technologies
Discuss the process for developing a national testing
algorithm
Explain how sensitivity, specificity, positive/negative
predictive value relate to development of an HIV rapid
testing algorithm
Explain the HIV rapid testing algorithm approved in your
country
Determine HIV status following a particular algorithm
Testing strategies and algorithms
Developing national testing algorithm
Measuring performance of HIV rapid tests
Interpreting HIV status
Handouts
Exercise #1: Interpreting HIV Status Using Testing Algorithm
Exercise #2: Interpreting HIV Status Using Testing Algorithm
Specify specific tests used in your country algorithm.
Notes on
Customization
Delete examples of HIV rapid tests not used in your country’s
algorithm and replace with in-country examples.
Replace testing algorithm diagram with country-specific
algorithm.
Module 4: HIV Testing Strategies and Algorithms
2005
1
Participant Manual
Strategies and
Algorithms
Strategies are defined as the testing approach used to meet a
specific need, such as for blood safety, surveillance, and
diagnosis. For a given strategy, multiple algorithms may be
used depending on the needs of testing settings.
Algorithms are defined as the combination and sequence of
specific tests used in a given strategy. The number of
algorithms should be limited.
HIV Testing
Strategies
Parallel and serial testing can be part of any testing strategy.
Parallel testing means samples are tested simultaneously by
two different tests.
Serial testing means samples tested by a first test. The results
of the first test determine whether additional testing is required.
Testing
Algorithms
Should be
Developed at
National Level
Before any test is adopted in country for use, a series of key
steps must be taken. These steps include:
•
Identifying appropriate tests
•
Developing an algorithm
•
Building consensus
•
Developing policy
•
Bringing into national scale
•
Reviewing testing algorithms annually
Because multiple tests are marketed and available in-country,
each country must identify the appropriate tests for use within
given environment. A standardized approach to developing an
algorithm must be taken. This involved building consensus and
developing policy before a test is brought to national scale.
After testing algorithms are adopted and implemented
nationally, these algorithms must be reviewed annually to
determine if test are performing as expected, and to determine if
any changes need to be made to the algorithm.
Timeline for
Developing
National Testing
Algorithm
A number of activities must be accomplished before a test is
fully implemented. Tests undergo a series of evaluations before
they are adopted for use country-wide. The timeline from
beginning of evaluation activities to the time when tests are fully
implemented is approximately 9 months to 1 year.
Module 4: HIV Testing Strategies and Algorithms
2005
2
Participant Manual
Timeline for Developing National Testing Algorithm
Phases
Key Activities
Timelines
Project Development
Determine Capacity
1 to 3 months
Literature Review
Situation Analysis
Needs Analysis
Test Evaluation /
Algorithm (Phase I)
Proposal
3 to 6 months
Ethical Review
Procurement
Establish Panels
Evaluation
Analysis of Data
Algorithm Decision
Publish Findings
Development (Phase II)
Site Selection
6 months
Training of Staff
On Site Evaluation
Algorithm Approval
Monitoring (Phase III)
Pilot Manuals
> 3 months
Monitor Performance
Publish Algorithm
The test evaluation process includes three phases:
Phase I Samples with known HIV results are tested at the
laboratory using a number of different kits. The results of the
rapid tests are compared with the original laboratory results.
Ones that worked the best are then evaluated in phase II.
Phase II Clients are tested at a rapid testing site by the tests
under evaluation, and the same sample is tested at the
laboratory using laboratory tests. The results are compared, and
the rapid tests that worked the best will be used in phase III.
Phase III Rapid tests that have been approved by the MOH are
used at rapid testing sites.
Module 4: HIV Testing Strategies and Algorithms
2005
3
Participant Manual
Advantages of
National Testing
Strategies and
Algorithms
Nationally adopted testing strategies and algorithms facilitates:
Country-level standardization of tests used in country supporting a limited number of tests is more feasible and
practical than many different tests.
Procurement and supply management - using
standardized tests allows for bulk procurement which
facilitates controlling costs.
Training –Implementation of a national training program
is eased when test sites follow the same testing
algorithm. This facilitates pre-planning of workshop, as
well as assuring staff who move from one test site to
another will not require total re-training.
Quality Assurance – national oversight of quality of
testing operations is easier when test sites use the same
tests and have similar operations.
Key Factors in
Determining a
Country’s
Algorithm
A number of factors contribute to the selection of specific tests
in a country’s algorithm. These factors include: test
performance in country, test availability in country, program
needs, ease of use, type of specimen, cost, and potential need
to differentiate between HIV 1 & HIV 2.
Evaluating Test
Performance:
Basic Terms
One of the factors that determine a country’s algorithm is the
performance the tests. This performance is based on how close
a test under consideration agrees 100% with the result of
another test, i.e., reference method or “gold standard”, also
referred to as the true result or actual HIV status.
Sensitivity (Se) and Specificity (Sp) relate to the performance
of the test capacity. Sensitivity of a test is its capacity to
correctly identify people who are infected with HIV
Positive and negative predictive value (PPV & NPV) relate to
the probability of the individual being infected or not infected
when the test yields a positive or negative result
Module 4: HIV Testing Strategies and Algorithms
2005
4
Participant Manual
Calculating
Sensitivity,
Specificity, PPV &
NPV
Actual HIV Status (Gold Standard)
Total Result
HIV Infected
HIV Uninfected
Total
Positive
A
B
A+B
Negative
C
D
C+D
A+C
B+D
Total
Sensitivity = A ÷ (A+C)
Positive Predictive Value = A ÷ (A+B)
Specificity = D ÷ (B+D)
Negative Predictive Value = D ÷ (C+D)
This table above provides the formulas for calculating Se, Sp,
PPV, & NPV.
A = true positives; that is the test under evaluation yields a
positive result AND the “gold standard (the reference test
method used to compare all other test results to)” yielded a
positive result.
C = false negatives; that is the test under evaluation yielded a
negative result, while the “gold standard” or true value was
positive.
A+C = all people who are truly infected with HIV
B = false positives; that is the test under evaluation yields a
positive result, while the “gold standard” or true value was
negative
D = true negatives; that is the test under evaluation yields a
negative result AND the “gold standard” or true value yielded a
negative result
B+D = all people who are truly NOT infected with HIV
Calculating
Sensitivity,
Specificity, PPV &
NPV (Cont’d)
Actual HIV Status (Gold Standard)
Total Result
HIV Infected
HIV Uninfected
Total
Positive
A (370)
B (2)
A+B (372)
Negative
C (4)
D (624)
C+D (628)
A+C (374)
B+D (626)
1000
Total
Sensitivity = A ÷ (A+C) = 370 ÷ 374 = 98.9%
Specificity = D ÷ (B+D) = 624 ÷ 626 = 99.7%
PPV = A ÷ (A+B) = 370 ÷ 372 = 99.5%
NPV = D ÷ (C+D) = 624 ÷ 628 = 99.4%
The table above illustrates the calculation of Se, Sp, PPV &
NPV.
Evaluation of a rapid test on a panel of specimens that have
been tested by the gold standard to determine actual HIV status
is shown to contain HIV antibodies to 374 samples and no HIV
antibodies to 626 samples.
Module 4: HIV Testing Strategies and Algorithms
2005
5
Participant Manual
Of the 374 serum samples that were antibody positive, the rapid
test classified 370 of the samples as positive.
Of the 626 samples that were HIV antibody negative by the gold
standard, 624 were classified by the rapid tests as not
containing HIV antibodies.
The results for Se, Sp, PPV, and NPV ranges from 98.9% to
99.7 %. This means:
The test performance of the HIV rapid test under
evaluation is nearly as good as the results of the “gold
standard” or true value.
The HIV test will yield highly accurate results
But, while these results indicated very good performance, you
must also keep in mind that the performance is influenced by
the makeup of the population.
HIV Rapid Test
Performance
As illustrated in the example above, no test is 100% sensitive or
100 % specific when compared to the “gold standard”. In other
words, no test will yield exactly the same result of the “gold
standard” test.
How Prevalence
Affects PPV &
NPV
If you know the sensitivity and specificity of the kit and the
prevalence of HIV in the population, you can theoretically
calculate the predictive value of the performed test. The formula
for calculating the predictive values of a test is as follows:
(Prevalence) (Se)
PPV= ___________________________________
(Prevalence) (Se) + (1- Prevalence) (1- Sp)
(1-Prevalence) (Sp)
NPV= ___________________________________
(1-Prevalence) (Sp + (Prevalence) (1- Se)
=
PPV for 10 % prevalence population:
(10/100) (98.9/100)
----------------------------------------------------------------------- = 97.3%
(10/100) (98.9/100) + (1- 10/100) (1- 99.7/100)
PPV for 1% prevalence population:
(1/100) (98.9/100)
= --------------------------------------------------------------------------- = 76.9%
(1/100) (98.9/100) + (1- 1/100) (1- 99.7/100)
Module 4: HIV Testing Strategies and Algorithms
2005
6
Participant Manual
In general, the higher the prevalence, the higher the PPV. In this
example there is a 97.3 % probability that if a person tests
positive then this individual is indeed infected with HIV.
Conversely, the lower the prevalence, the lower the PPV. In the
example above, there is a 76.9% probability that if a person
tests negative, then this individual is indeed not infected with
HIV.
Testing Algorithm Testing Algorithm describes the sequence of tests to be
Describes the
performed. An HIV Positive Status should be based upon the
Sequence of Tests outcome of 2 or more tests.
to be Performed
When two test results disagree (one is reactive, the other nonreactive), the finding is called “discordant.” In this case, a third
test must be performed.
Always follow the sequence of the tests in the algorithm.
Ideal Algorithm
The ideal algorithm used in one in which tests are highly
sensitive and highly specific.
Both tests should not share the same false negatives and false
positives, and a third test is performed, if needed.
Parallel Testing
Algorithm
In a parallel testing algorithm, two different tests are performed
at the same time. If a 3rd test or tiebreaker is required, this test
is different than test 1 or test 2.
Blood
Blood Sample
Sample
Test 2
Test 1
Both
Both
Reactive
Reactive
Both
Both
Non-reactive
Non-reactive
Report
Report
Positive
Positive
Report
Report
Negative
Negative
Discordant
Discordant
Result
Result
Test
Test 33
Module 4: HIV Testing Strategies and Algorithms
2005
7
Reactive
Reactive
Result
Result
Non-reactive
Non-reactive
Result
Result
Report
Report
Positive
Positive
Report
Report
Negative
Negative
Participant Manual
Exercises
•
When both tests are reactive, the final HIV result is
positive.
•
When both tests are non-reactive, then the final HIV
result is negative.
•
If one test is reactive and the other is non-reactive, then
a third test known as a tiebreaker is performed.
•
The tie-breaker determines the final result – if the
tiebreaker is reactive, then the final HIV result is positive;
if the tiebreaker is non-reactive, then the final HIV result
is negative. The test used as the tiebreaker is a different
test than test 1 or 2.
Interpreting HIV Status Using Testing Algorithm
At the end of this module, you will find two exercise sheets.
Read the instructions and write your answers in the space
provided.
Possible HIV Test
Outcomes:
Parallel Testing
Algorithm
There are a variety of combinations of outcomes when following
a parallel testing algorithm.
Parallel Testing Algorithm
TEST 1
TEST 2
HIV Status
Non-reactive
Non-reactive
Negative
Reactive
Reactive
Positive
Non-reactive
Reactive
Non-reactive
Negative
Reactive
Non-reactive
Non-reactive
Negative
Non-reactive
Reactive
Reactive
Positive
Reactive
Non-reactive
Reactive
Positive
Before any test is adopted in country for use, a series of
key steps must be taken to evaluate the tests before
they are fully adopted for use country-wide.
The ideal algorithm used in one in which tests are highly
sensitive and highly specific.
No test is 100% sensitive or 100 % specific when
compared to the “gold standard”.
Always follow the sequence of the tests in the algorithm.
Key message
Module 4: HIV Testing Strategies and Algorithms
2005
TEST 3
8
Participant Manual
Module Review
Find out how much you have learned by answering these questions.
Explain the importance of a tests: ’ Se, Sp, PPV, NPV
- Se ______________________________________________________________
- Sp ______________________________________________________________
- PPV _____________________________________________________________
- NPV _____________________________________________________________
Explain the testing algorithm adopted by MOH. What rapid tests are used and
in what order?
__________________________________________________________________
__________________________________________________________________
__________________________________________________________________
__________________________________________________________________
__________________________________________________________________
Module 4: HIV Testing Strategies and Algorithms
2005
9
Participant Manual
Exercise #1: Interpreting HIV Status Using Parallel Algorithm
A client has come in for an HIV Rapid Test. Following the parallel algorithm, you conducted
three tests. The results are shown below.
Interpret the HIV Rapid Test results for Test #1, Test #2, and Test #3
Decide HIV Status
Test #1
Test #2
Test #3
Result: _________________
Result: _________________
Result: _______________
Final HIV Status________________________
Module 4: HIV Testing Strategies and Algorithms
2005
11
Participant Manual
Exercise #2: Interpreting HIV Status Using Serial Algorithm
A client came in for an HIV Rapid Test. Following the serial algorithm, you conducted two
tests. The results are shown below.
Interpret the HIV Rapid Test results for Test #1 and Test #2.
Decide if a third test is needed. If not, what is the client’s HIV status?
Test #1
Test #2
Result: _________________
Result: _________________
Is a third test needed? ___Yes ___No
If no, what is the HIV Status? _______________
Module 4: HIV Testing Strategies and Algorithms
2005
12
Participant Manual